广东水利水电Issue(2)：1-5,31,6.
基于长短期记忆网络的连江流域径流模拟
Runoff Simulation in the Lianjiang River Basin Based on the Long Short-Term Memory
摘要
Abstract
The construction of water conservancy project and flood prevention are supported by runoff simulation under climate change as a scientific basis.In order to predict the runoff process in heavy rainfall and flood-prone watersheds,this paper constructed a model to simulate runoff based on the long short-term memory network(LSTM),selected the model inputs,structure and parameters,and simulated and predicted the runoff process in Lianjiang River Basin from 2011 to 2020.The results show that:the LSTM model had high accuracy when simulating runoff in Lianjiang River basin,with NSE and R2 reaching 0.88 and 0.90 in the test period,respectively;the LSTM model was able to simulate the small flood peaks in the flood season and predict the main flood peaks in the flood season,which is suitable for runoff simulation of mountainous rivers with large runoff variability;and the LSTM model had strong generalization,and it can predict the future runoff steadily with a low degree of overfitting.The research results aim to improve the accuracy of runoff simulation and prediction in mountainous rivers,and increase the capacity of prevention and control of flood and drought disasters in the basin.关键词
径流模拟/LSTM/深度学习模型/连江流域/山溪性河流Key words
runoff simulation/LSTM/deep learning model/Lianjiang River basin/mountainous rivers分类
天文与地球科学引用本文复制引用
陈鑫池,徐嘉远,黄东,李亚泽..基于长短期记忆网络的连江流域径流模拟[J].广东水利水电,2025,(2):1-5,31,6.基金项目
2023年广东省科技创新战略专项资金科研项目(编号:sky2023-05). （编号:sky2023-05）
